Quote:
Originally Posted by layi
PLEASE CAN ANYONE ASSIST ME IN THE BUILDING OF A TRADING SYSTEM BASED ON ARTIFICIAL NEURAL NETWORK .THE OBJECTIVE IS TO PREDICT THE MOMENT WHEN THE EUR/USD RATE WOULD INCREASE OR DECREASE ON AVERAGE OF ATLEAST 5 PIPS IN THE NEXT 75 MINUTES (FIVE TIME UNITS OF 15 MINUTES EACH).
THE INPUT OF THE NEURAL NETWORK WILL INTENT TO DESCRIBE EACH TRADING DAY IN TERMS OF PAST EVOLUTION OF THE EUR/USD OVER A 75 MINUTES PERIOD. THE VALUE WILL BE BASED ON EUR/USD CURRENT CLOSE,IT’S MOVING AVERAGE OVER 75 MINUTES,MINIMUM AND MAXIMUM VALUE OVER THE SAME TIME INTERVAL.
THE VALUE FOR THE OUTPUT ATTRIBUTE ASSIGNED WILL BE:
A) “+1” IF THE EXCHANGE RATE’S MOVING AVERAGE FOR THE NEXT 75 MINUTES EXCEEDED THE CURRENT CLOSE VALUE BY MORE THAN FIVE PIPS (THAT IS THE NETWORK GENERATED A BUY SIGNAL)
B) “+1” IF THE EUR/USD MOVING AVERAGE FOR THE NEXT 75 MINUTES FELL MORE THAN FIVE PIPS COMPARED TO THE CURRENT CLOSE VALUE (THAT IS THE NETWORK WILL GENERATE A SELL SIGNAL)
C) “0”IF THE RATE’S MOVING AVERAGE FOR THE NEXT 75 MINUTES OSCILLATED BETWEEN PLUS/MINUS FIVE PIPS COMPARED TO THE CURRENT CLOSE VALUE (THAT IS THE NETWORK DID NOT GENERATE ANY SIGNAL)
IN CHOOSING THE OPTIMAL ARCHITECTURE FOR THE NEURAL NETWORK WILL BE TWO INPUTS,TWO HIDDEN LAYERS WITH FOUR AND TWO NODES EACH AND ONE OUTPUT  |
i want to ask who wrote this, how can you tell us the optimal ARCHITECTURE!!!! do you really read about ann....if you do that find there is no way to find best ARCHITECTURE.... it is practical and depend on you variable ......study more and do not think about this ways....ann is more difiicalt than you think .....